Pointwise simultaneous approximation by combinations of Bernstein operators (Q2575614)

For \(f\in C[0,1]\), let \[ B_n(f,x)=\sum _{k=0}^{n}p_{n,k}(x)f(k/n),\quad p_{n,k}(x)={n\choose k}x^k(1-x)^{n-k} \] be the classical Bernstein operator and let \(B_n(f,r,x)\) be the combinations of them defined by \textit{Z. Ditzian} and \textit{V. Totik} [Moduli of smoothness. Springer Series in Computational Mathematics, 9. New York etc.: Springer-Verlag (1987; Zbl 0666.41001)] as \[ B_n(f,r,x)=\sum_{i=0}^{r-1}C_i(n)B_{n_{i}}(f,x), \] where \(n_i\) and \(C_i(n)\) satisfy the following conditions: \[ n=n_0<\cdots <n_{r-1}\leq An;\quad \quad \sum_{i=0}^{r-1}| C_i(n)| \leq C; \] \[ \sum_{i=0}^{r-1}C_i(n)=1;\quad \quad \sum_{i=0}^{r-1}C_i(n)n_i^{-\rho }=0,\;\rho =1,2,\dots, r-1. \] Pointwise direct and inverse approximation results by these combinations were proved by \textit{S. Guo, C. Li, X. Liu} and \textit{Z. Song} [J. Approximation Theory 107, No. 1, 109--120 (2000; Zbl 0967.41011)] and \textit{L. S. Xie} and \textit{T. F. Xie} [Chin. Ann. Math., Ser. A 20, No. 4, 467--478 (1999; Zbl 0949.41012)] using the \(2r\)th Ditzian-Totik modulus of smoothness, defined as \[ \omega _{\varphi ^\lambda }^{2r}(f,t)=\sup_{0<h\leq t }\sup_{0\leq x\leq 1 }| \triangle_{h\varphi^{\lambda}(x)}^{2r}f(x)| , \] where \(0\leq \lambda \leq 1\), \(\varphi (x)=(x(1-x))^{1/2}\), \(\triangle _t^{2r}f(x)=\sum_{j=0}^{2r}(-1)^j{2r\choose j}f(x+(r-j)t)\) when \(rt\leq x\leq 1-rt\) and \(\triangle _t^{2r}f(x)=0\) otherwise. In the paper the author proves new direct and inverse results on pointwise simultaneous approximation by the derivatives of these combinations of Bernstein operators. The main result is Theorem. Let \(r,s\in N\), \(0\leq \lambda \leq 1\) and \(s<\alpha <s+2r/(2-\lambda)\). Then for all \(f^{(s)}\in C[0,1]\) \[ \begin{aligned} {n^s(n-s)!\over n!}B_n^{(s)}(f,r,x)-f^{(s)}(x)&= O\left(\left(n^{-1/2}\delta_n^{1-\lambda }(x)\right)^{\alpha -s}\right)\;as \;n\rightarrow \infty\\ \Leftrightarrow \omega _{\varphi ^\lambda }^{2r}(f^{(s)},t)&= O(t^{\alpha -s})\;as \;t\rightarrow +0, \end{aligned} \] where \(\delta _n(x)=\varphi (x)+n^{-1/2}\).